Decision support and human oversight
Use AI to compare options while a person owns the decision.
In this lesson · F6 of 8
All lessons in this track →Progress for this track is saved in this browser only. No completion badge is issued. View your learning.
Start with the idea
Use AI to compare options while a person owns the decision.
Decide what an option must satisfy before ranking it. A large benefit cannot cancel a required condition. AI can help compare options; the responsible person still makes the decision.
Required condition: Something an option must satisfy to be considered.
See the example and check your understanding
A made-up teaching example
Room A has 20 seats and confirmed accessibility. Room B has 30 seats, but its accessibility is unknown. Twenty people are expected, and confirmed accessibility is required. Extra seats do not fix that missing evidence.
Ask yourself
Can Room B win just because it has more seats?
Show the explanation
No. It does not yet meet the evidence requirement. Room A fits the supplied conditions; Lee still owns the final choice.
Start with these checks
- Check required conditions before scores.
- Do not treat unknown as acceptable.
- Keep the final decision with its named owner.
Use the different practice case below and check every requirement in its full instructions.
The video explains the main idea. Use the sources, instructions and checks below to complete the practice.
Captions appear here during playback.
Decision support and human oversight
Use AI to compare options while a person owns the decision.
Video playback is unavailable. Read the descriptive transcript below, go directly to practice, or open the video file.
Source-based teaching example · Fictional data · Reviewed 2026-10-03. Microsoft Mark · locally synthesized. On a small screen, use fullscreen for diagram details or read the transcript.
Read the descriptive transcript
0:00 · What you will learn
Use AI to compare options while a person owns the decision.
What appears on screen
Use AI to compare options while a person owns the decision.
0:06 · The idea
Decide what an option must satisfy before ranking it. A large benefit cannot cancel a required condition. AI can help compare options; the responsible person still makes the decision.
What appears on screen
Decide what an option must satisfy before ranking it. A large benefit cannot cancel a required condition. AI can help compare options; the responsible person still makes the decision.
0:21 · One useful term
Required condition. Something an option must satisfy to be considered.
What appears on screen
Something an option must satisfy to be considered.
0:29 · A worked example · 1
Room A has 20 seats and confirmed accessibility. Room B has 30 seats, but its accessibility is unknown. Twenty people are expected, and confirmed accessibility is required.
What appears on screen
Room A has 20 seats and confirmed accessibility. Room B has 30 seats, but its accessibility is unknown. Twenty people are expected, and confirmed accessibility is required.
0:45 · A worked example · 2
Extra seats do not fix that missing evidence.
What appears on screen
Extra seats do not fix that missing evidence.
0:51 · Your turn to think
Can Room B win just because it has more seats? Pause and decide before continuing.
What appears on screen
Can Room B win just because it has more seats? Pause the video if you need more time.
1:07 · Why this answer works
No. It does not yet meet the evidence requirement. Room A fits the supplied conditions; Lee still owns the final choice.
What appears on screen
No. It does not yet meet the evidence requirement. Room A fits the supplied conditions; Lee still owns the final choice.
1:19 · Check 1
Check required conditions before scores.
What appears on screen
Check required conditions before scores.
1:24 · Check 2
Do not treat unknown as acceptable.
What appears on screen
Do not treat unknown as acceptable.
1:29 · Check 3
Keep the final decision with its named owner.
What appears on screen
Keep the final decision with its named owner.
1:35 · Now try a different case
Open the lesson and try its different practice case. Use every check in the full task. Watching alone does not mark the lesson complete.
What appears on screen
Open the lesson Try the new case Check and revise The full task and its requirements are in the linked lesson.
Watching explains the example. Complete the lesson practice to check what you can do. Explore all visual lessons →
Before you start
Prerequisite: F4–F5 claim checks and bounded briefs. Offline/no-cost route: paper, a notes editor and the synthetic packet below. Optional tools must already be authorized. No accounts, paid calls, workplace records or external actions are required.
What you will be able to do
- Decompose a decision and assign owners.
- Identify automation bias, missing stakeholders, and second-order effects.
- Design meaningful review and contestability.
Part 1 Support the decision without taking its authority
Frame the problem, criteria and hard constraints before requesting a ranking. Generate options, then evaluate evidence and tradeoffs. Meaningful human oversight requires an authorized reviewer with competence, time, relevant information and the ability to stop or change the result. A ceremonial click is insufficient. Write a human baseline before viewing AI recommendations to reduce anchoring. Include affected-party consequences, dissent and a way to challenge errors. AI can support deliberation but cannot inherit reserved authority.
Part 2 Worked synthetic example
Synthetic workshop options: Room A seats 20, accessible, available 14 October. Room B seats 30, accessibility unknown, available 14 October. Twenty participants are expected; confirmed accessibility is a hard requirement. A model picks B for spare capacity. The reviewer rejects the ranking because B fails the evidence requirement. A fits supplied constraints; final selection remains with fictional coordinator Lee. No booking occurs.
Part 3 Guided practice and feedback
Independently rank before reading the recommendation. List capacity, date and accessibility. If expected attendance rises to 25, A no longer fits and B is still unresolved: seek an accessible alternative or revised attendance plan. Do not relax the hard requirement silently.
Review and next step
Self-review demonstrates practice, not independent certification or operational approval. F7 adds connected tools and action boundaries to the decision.
Substantive review: 2026-10-02. See the configuration guide. OWASP excessive-agency guidance discusses the risks of unnecessary capabilities, permissions and autonomy.